Skip to content
Olivares AI docs

Olivares AI documentation

Run AI tools and sessions on your own infrastructure.
Start here Find your path

Start with a tutorial, follow a guide for a task, or look up an interface in the reference.

Tutorials Run the binary and build a read/write access graph.
How-to guides Connect a source, deploy, verify a release or forward audit to Splunk.
Reference REST API, event bus (AsyncAPI), modules, CLI and configuration.
Explanation Architecture, access graph, security, threat model and licensing.

What is this?

Section titled “What is this?”

Olivares AI runs AI tools such as Claude Code, Codex and Grok Build with shared sessions, resource access, permissions, policies, budgets and audit evidence. The binary includes the console and runs on your own infrastructure. The access map compares permitted access with observed reads and writes.

Olivares AI has no mandatory telemetry. Your configuration selects model API calls, SIEM and webhook outputs, and any external embedding provider. Some capabilities are partial or planned; check their documented limits.

Limits

Check the status and limits of each capability before using it. See Honesty & limits.

Open core

The product (engine, modules, web) is AGPL-3.0-only; the SDK and connectors are Apache-2.0; the Business and Enterprise modules are commercial. See Open core & licensing.

Now: the approvals waiting for you with what each one will run, recent sessions across your tools, spend and each account's usage limits.
Now — the approvals waiting for you with what each one will run, recent sessions across your tools, spend and each account's usage limits.
Cockpit: what every agent is doing now across your workspaces, with cost, usage limits per account, the machines reporting, approvals to decide in place and what changed today.
Cockpit — what every agent is doing now across your workspaces, with cost, usage limits per account, the machines reporting, approvals to decide in place and what changed today.
Session: a running session: the conversation beside its usage limits, context, agents, background tasks, plan and changed files.
Session — a running session: the conversation beside its usage limits, context, agents, background tasks, plan and changed files.
AI tools: which tools are installed, the accounts signed in to each with their plan, and one action to sign in the account that still needs it.
AI tools — which tools are installed, the accounts signed in to each with their plan, and one action to sign in the account that still needs it.
Sign in a tool: Claude Code signs in with its own sign-in page: open it, paste back the code it shows, or run the same step from a terminal. Olivares never sees the password.
Sign in a tool — Claude Code signs in with its own sign-in page: open it, paste back the code it shows, or run the same step from a terminal. Olivares never sees the password.
Identity and access: the user sources people sign in from, every user with their groups and role, and the groups mapped to roles.
Identity and access — the user sources people sign in from, every user with their groups and role, and the groups mapped to roles.
Permissions: what each role may do, and the policies that are on with the rules they hold.
Permissions — what each role may do, and the policies that are on with the rules they hold.
Access map: what each agent reads and writes, checked against what it was granted, with the one write nobody granted flagged for review.
Access map — what each agent reads and writes, checked against what it was granted, with the one write nobody granted flagged for review.
Budgets and spend: spend this month, per day and per tool, and the budgets that alert, slow down or block, one of them over its limit.
Budgets and spend — spend this month, per day and per tool, and the budgets that alert, slow down or block, one of them over its limit.
Audit: signed records of approvals, allowed and denied tool calls, sign-ins and policy changes.
Audit — signed records of approvals, allowed and denied tool calls, sign-ins and policy changes.
Integrations: connected MCP servers with their access mode, allowed tools and the accounts that use them, plus the evidence, alert and event connectors.
Integrations — connected MCP servers with their access mode, allowed tools and the accounts that use them, plus the evidence, alert and event connectors.
Executive overview: spend and sessions over 30 days, spend by tool, where agents run, compliance evidence and risk by agent.
Executive overview — spend and sessions over 30 days, spend by tool, where agents run, compliance evidence and risk by agent.
Parallel sessions: six live sessions side by side, each in its own worktree branch, with its latest output, changed files, cost and usage limit, and one waiting for your approval.
Parallel sessions — six live sessions side by side, each in its own worktree branch, with its latest output, changed files, cost and usage limit, and one waiting for your approval.